Pediatrics World Conference 2026

Abstract

Postoperative complications in pediatric surgery remain a significant cause of morbidity, prolonged hospitalization, and healthcare resource strain. Early identification of high-risk patients is crucial to improve outcomes and optimize care. We developed an Artificial Intelligence (AI) system integrating Machine Learning (ML) models to predict pediatric postoperative complications in real time, focusing on high-risk procedures including hepatopancreatobiliary, cardiac, and abdominal surgeries at Medical Center in Mexico. 
A retrospective-prospective study was conducted with 500 pediatric surgical patients at Medical Center in Mexico. Clinical data collected included demographics, comorbidities, BMI, preoperative labs, surgical details, operative time, and intraoperative blood loss. ML models (Random Forest, XGBoost, Neural Networks) were trained to predict infections, readmissions, sepsis, and procedure-specific adverse events. Model performance was evaluated using AUC, sensitivity, specificity, and precision. Feature importance analyses identified variables most predictive of outcomes.

Preliminary results show high predictive accuracy (AUC > 0.87), with operative time, preoperative albumin, BMI, and intraoperative blood loss as top predictors. Early AI-driven risk stratification enables identification of patients requiring intensive postoperative monitoring, personalized perioperative planning, and optimized resource allocation.
AI-powered predictive systems represent a breakthrough in precision pediatric surgery, enabling clinicians to make informed decisions, improve recovery, and reduce complications. Future work includes real-time integration in surgical workflows and prospective validation at Medical Center across diverse pediatric populations.